A novel partial volume effects correction technique integrating deconvolution associated with denoising within the PET image reconstruction process

摘要

Partial Volume Effects (PVE) represent a major source of degradation in PET imaging, introducing large biases especially for small structures. To correct for these effects, numerous methods have been proposed, working on reconstructed images or incorporated within the reconstruction process. For the latest case, most existing studies focus on the accurate determination of the system point spread function (PSF), using either direct physical measurements [1] or analytical [2] and Monte Carlo based estimations [3]. Within this implementation convolution by the PSF is performed either at the projection step only [4], or at both projection and backprojection steps [3]. Within this context, Rizzo et al. proposed an alternative approach with the use of the Lucy-Richardson (LR) deconvolution after each global iteration of the OSEM algorithm [5]. Their results on simulated and acquired data show improvements in image spatial resolution in comparison to conventional OSEM algorithm. However, a noise increase has been also observed visually but not quantitatively evaluated. This is not surprising since iterative deconvolution algorithms are known to significantly amplify noise. Methods that have been proposed to reduce noise propagation within the context of image based deconvolution include the use of wavelet based denoising [6]. This approach has shown significant noise reduction without loss of resolution recovery.